This is going to be my first ranking of adjusted margin of victory for the B1G this year. Normally, I would wait until closer to the midway point of the season, on the assumption that you can't learn anything valuable from 2 or 3 games, but I'm doing this very early this year as an experiment, to see if that assumption holds up. I explained the math on this last year, so even though I'll enjoy doing it again (since I'm a math geek), I'm going to save all the math for the end of the post, so it's easy for people to skip who don't care about it.
What these rankings represent is essentially a Sagarin or KenPom style value for teams, but based only on in-conference games. There are no assumptions built in, preseason or otherwise. They simply measure how teams have actually performed in the conference, relative to their opponents. The actual number is supposed to represent how much a team would expect to win or lose by on a neutral court against a perfectly average B1G team. They are:
Not a lot of surprises. The computer doesn't like MSU's performance so far, and IU seems a little low for a 3-0 team, while Illinois seems a little high for an 0-2 squad, but by and large, they seem to make sense.
As for IU specifically, the computer puts the Hoosiers right smack dab in the middle largely because, even though we overperformed slightly against Nebraska and Wiscy (2.6 and 1.7 points, respectively), we underperformed by 4.8 points against Rutgers. If the Rutgers performance was an aberration, you'll see IU gradually move up this list.
Early in the season, the disparity in schedule strength is obviously glaring from team to team. This accounts for Illinois ranking so far ahead of IU. Illinois has had the #1 SOS so far, while IU has had the #14 SOS. Based on the schedule remaining, IU is predicted to finish 10-8, and in 7th place.
Home court advantage has been minor in the early going, exactly 1 point. Last year, the final home court advantage was 2.96. Also, if last year is any indication, at least some of the teams at the very top and bottom will come back to the pack. Wiscy was the only double-digit team last year, at 11.12. Rutgers was the only double-digit at the bottom, at -11.78. Every other team was between 6 and -6.
Finally, since the whole reason I started doing this was as a gambling experiment, for tonight's games, Maryland should be a 20-point favorite, while Northwestern should be 4-point dogs. Last I saw, Maryland was favored by 22, while Northwestern was actually a 2-point favorite. My computer says betting on the Buckeyes tonight is where you can make some money.
The Math
For those who are interested in the math, this started many years ago as an experiment to come up with a pure alternative to Sagarin ratings that was based only on actual margin of victory. What I did was take each game, and adjust the real MOV according to the average MOVs of the participating teams. I then refigured each team's average MOV using the new numbers, and went back and forth through several iterations. I quickly realized that I only had to do a few iterations for the numbers to start trending toward a stable average. However, while the ratings were very good at predicting games in late December, once the conference season started, they fell apart. I figure the wildly unbalanced non-conference schedules make a lot of early games pretty useless for predicting conference success. Therefore, I now simply skip the non-con season, and use conference games only. I only do the B1G, because I don't care about other conferences, and I have much more limited time than I did back then.
These rankings are actually the averages of two different methods of trying to figure the same thing. One is a multiple-iteration approach, as described above. The other is a simulated approach, based on the RPI formula, which assigns different weights to opponent's strength, opponent's opponent's strength, etc. This is only my second year with this method, but I believe in the long run, I will find it to be more accurate than the original.
What these rankings represent is essentially a Sagarin or KenPom style value for teams, but based only on in-conference games. There are no assumptions built in, preseason or otherwise. They simply measure how teams have actually performed in the conference, relative to their opponents. The actual number is supposed to represent how much a team would expect to win or lose by on a neutral court against a perfectly average B1G team. They are:
Code:
Rank Team Avg
1 UM 14.74
2 Iowa 7.61
3 Mary 6.81
4 PU 4.99
5 Ill 3.43
6 Wiscy 2.45
7 OSU 2.44
8 Ind 0.72
9 PSU -2.25
10 MSU -2.38
11 NW -2.64
12 Neb -7.68
13 Minny -11.06
14 RU -12.07As for IU specifically, the computer puts the Hoosiers right smack dab in the middle largely because, even though we overperformed slightly against Nebraska and Wiscy (2.6 and 1.7 points, respectively), we underperformed by 4.8 points against Rutgers. If the Rutgers performance was an aberration, you'll see IU gradually move up this list.
Early in the season, the disparity in schedule strength is obviously glaring from team to team. This accounts for Illinois ranking so far ahead of IU. Illinois has had the #1 SOS so far, while IU has had the #14 SOS. Based on the schedule remaining, IU is predicted to finish 10-8, and in 7th place.
Home court advantage has been minor in the early going, exactly 1 point. Last year, the final home court advantage was 2.96. Also, if last year is any indication, at least some of the teams at the very top and bottom will come back to the pack. Wiscy was the only double-digit team last year, at 11.12. Rutgers was the only double-digit at the bottom, at -11.78. Every other team was between 6 and -6.
Finally, since the whole reason I started doing this was as a gambling experiment, for tonight's games, Maryland should be a 20-point favorite, while Northwestern should be 4-point dogs. Last I saw, Maryland was favored by 22, while Northwestern was actually a 2-point favorite. My computer says betting on the Buckeyes tonight is where you can make some money.
The Math
For those who are interested in the math, this started many years ago as an experiment to come up with a pure alternative to Sagarin ratings that was based only on actual margin of victory. What I did was take each game, and adjust the real MOV according to the average MOVs of the participating teams. I then refigured each team's average MOV using the new numbers, and went back and forth through several iterations. I quickly realized that I only had to do a few iterations for the numbers to start trending toward a stable average. However, while the ratings were very good at predicting games in late December, once the conference season started, they fell apart. I figure the wildly unbalanced non-conference schedules make a lot of early games pretty useless for predicting conference success. Therefore, I now simply skip the non-con season, and use conference games only. I only do the B1G, because I don't care about other conferences, and I have much more limited time than I did back then.
These rankings are actually the averages of two different methods of trying to figure the same thing. One is a multiple-iteration approach, as described above. The other is a simulated approach, based on the RPI formula, which assigns different weights to opponent's strength, opponent's opponent's strength, etc. This is only my second year with this method, but I believe in the long run, I will find it to be more accurate than the original.
